The Trade Desk review — excellent at its job, and its job is not Amazon retail

What it is, plainly, and which DSP we are talking about

Two disambiguations first, because this page will be read by Amazon sellers and both of them cause real confusion.

A DSP is a Demand-Side Platform — software for buying display, video and audio inventory programmatically. It is not Amazon's Delivery Service Partner programme, which is the parcel-delivery franchise. Different business entirely.

The Trade Desk is not Amazon DSP. They are two separate platforms. Amazon DSP is Amazon's own buying system with Amazon shopping signal behind the targeting. The Trade Desk is independent — it describes itself on its own homepage as "The leading independent DSP built for data-driven marketers". A brand can use both, and many large ones do, but they are not substitutes and pricing them against each other is usually the wrong comparison.

What The Trade Desk sells is a self-service platform for buyers: campaign management, bidding, identity resolution, measurement integrations and access to inventory across the open internet, connected TV, audio, digital out-of-home and retail media networks. It does not own media, which is the single most important structural fact about it and the source of most of its strengths.

Scale, from its own fiscal 2025 results release: $2.896 billion of revenue, up 18% year over year, on $13.4 billion of gross spend transacted across the platform, with GAAP net income of $443.3 million. Whatever else is true, this is a large, profitable, established platform, and nothing in this review should be read as doubting that.

What it is genuinely excellent at

Four things, and a buyer who uses it would recognise all four.

Independence. The Trade Desk does not own the inventory it buys. That removes a structural conflict that exists everywhere a platform is also a seller of media, and it is the reason a great many agencies chose it in the first place. When you ask it where to spend, there is no house inventory it is quietly steering you toward.

Connected TV. This is where the platform has invested hardest and where its reach and measurement work are strongest. If your growth question involves streaming inventory at national scale, this is a first-choice platform rather than a compromise.

Supply-path work. The Trade Desk has done public, sustained work on making the route between buyer and publisher legible — which exchanges take what, and which paths are worth using. Most buyers benefit from that without ever thinking about it. It is unglamorous engineering and it is real.

Retention. Their fiscal 2025 release states customer retention "over 95% during the year, as it has for the past twelve consecutive years". Twelve consecutive years of that is a difficult thing to fake and a difficult thing to achieve. Ask which retention it is — logo or revenue — because those differ, but take the claim seriously either way.

Add one more that is not a feature: it is a public company. Everything above is checkable against filings rather than accepted from a sales deck. Almost nothing else in the DSP or retail media category can be verified that way, including us.

What being public tells you that no private vendor's page can

This is the most useful and least-used part of reviewing The Trade Desk, and it applies to no other vendor most Amazon brands will shortlist.

From the same fiscal 2025 release: full-year revenue grew 18%, from $2.445 billion to $2.896 billion. Fourth-quarter revenue was $846.8 million against $741 million a year earlier — 14% growth. So the exit rate was slower than the full-year rate.

We are not going to editorialise about that, and neither should you on the strength of two data points. What matters for a buyer is the principle: a public vendor hands you a trend line, and a private one hands you a case study. When you are deciding whether to build a two-year dependency on a platform, being able to read the direction of travel from an audited filing is a genuine advantage of choosing a listed company — and its absence is a genuine unknown when choosing anyone else.

Apply that symmetrically. We are private. So are Quartile, Pacvue, Intentwise, Criteo's retail media arm and the great majority of this category. None of us can be checked the way The Trade Desk can. The substitute is a reference call with a customer of similar size who has been with them for more than two years, and a direct question about what happens to your rate and your team if the company changes hands.

Do the same arithmetic on scale that nobody bothers with, because it puts the platform in proportion. Amazon's own DSP page states that its managed-service option "typically requires a minimum spend of $50,000". The Trade Desk transacted $13.4 billion in 2025 — roughly 268,000 times that floor in a single year. This is not a boutique, and you should not expect boutique attention at a boutique budget.

The honest limitation for an Amazon-first advertiser

Everything above is a reason to like the platform. Here is where it stops being the right tool, stated without any suggestion that it is a bad one.

The Trade Desk is not the route into Amazon's retail signal. It does not give you Amazon's shopping behaviour as targeting input, and it does not give you Amazon Marketing Cloud, the clean room where event-level data from Amazon DSP and Amazon sponsored ads sits together. Amazon's own DSP page describes that capability in its own words: you can "perform custom analysis on your Amazon DSP campaigns using event-level datasets in Amazon Marketing Cloud".

Why that matters more than any feature comparison: if you sell primarily on Amazon, your display advertising and your sponsored ads are competing for credit on the same orders. Two dashboards will each claim the sale, and an account reading both will conclude that both worked. The only place to settle it is a clean room with both event streams in it.

So the sentence to hold onto is narrow and fair: The Trade Desk is excellent at buying reach, and it is not where an Amazon-first brand reconciles its Amazon advertising. Those are different jobs, and a brand that needs both may legitimately buy both.

One more limitation that is about fit rather than capability: this is a platform, not a service. Somebody has to sit in it daily — building the campaigns, watching the supply paths, managing the audiences. If your organisation does not have that person, the platform's quality is not the constraint. The staffing is.

That last point deserves a number rather than an assertion, because it is the one most likely to be waved through in a business case. Amazon's own DSP page states that its managed-service option "typically requires a minimum spend of $50,000", varying by country, which is a reasonable proxy for the budget level at which programmatic starts being a real job rather than an experiment. At that level you are running $600,000 a year of media. A trader who can genuinely operate an independent demand-side platform — build the campaigns, manage the audiences, read the supply paths, keep the pacing honest — is a specialist hire, and the fully-loaded cost of one is a material fraction of that media budget before a single impression is bought.

The honest conclusion is not that platforms are bad value. It is that the platform decision and the staffing decision are the same decision, and evaluating the first without pricing the second is how organisations end up with an excellent console nobody has time to open. Whichever way you go, write the name of the person who will operate it at the top of the business case. If the box is empty, you are shopping in the wrong category and should be looking at managed offers instead.

The 79.9% problem — the arithmetic that decides whether you need the clean room

Here is the calculation that should sit at the centre of any DSP review and never does.

Across 30 advertisers in July 2026, our live Amazon DSP API pull recorded 57,137 attributed purchases, of which 20.1% were new to the brand. Turn that around: 79.9% of attributed purchases came from shoppers who already knew the brand.

That share is where the whole incrementality question lives. A new-to-brand purchase is comparatively easy to believe in — the shopper had not bought from you before, and something introduced them. An existing-customer purchase attributed to a display impression is a much harder claim, because a meaningful proportion of those people were going to buy anyway. Last-click attribution cannot separate the two. It sees a click or a view followed by an order and assigns credit; it has no way to represent the counterfactual, because the counterfactual never happened.

So the practical question for any display budget is: what fraction of the credited orders sits in that 79.9%, and how much of it would have arrived without the media? On a $100,000-a-month display budget, getting that fraction wrong by twenty points is $240,000 a year of re-authorised spend resting on an assumption.

Holdouts and matched controls answer it. Withhold the audience from a randomised slice, or match a comparable untreated group, and measure the difference in outcomes rather than the outcomes themselves. That requires the retail events and the ad events in one place — which is what Amazon Marketing Cloud is for, and which is the reason platform choice and measurement choice are the same decision for an Amazon-first brand.

Be fair about who else does this. Criteo advertises incrementality testing on its own advertiser pages. Tinuiti runs a published incrementality practice. Skai lists incrementality testing on its rate card from Enterprise Premier upward. This is not a capability we invented and we are not going to imply anyone lacks it. The question that separates vendors is who designs the test, who computes it, and whether it is written into the statement of work — and that question applies to us exactly as much as to them.

How to read reviews of a platform that keeps changing

A structural warning that applies across this whole category and is worth more than any individual star rating.

Date the review before you read it. Retail media and programmatic tooling has been rebuilt repeatedly in the last three years — products absorbed into suites, platforms renamed, AI layers added, ownership changed. A review written two years ago frequently describes an interface and a feature set that no longer exists. That is true of The Trade Desk, and it is true of nearly every alternative you will compare it against.

We will not quote star ratings or review counts here at all. The major review sites block automated reading, and search-result snippets of those pages have contradicted the live pages more than once in our own research. Printing a number we could not open ourselves would be the same error we are warning you about.

What to do instead, in four steps:

• Get two reference calls with customers of roughly your size and spend, one of them at least two years in, and ask what has changed since they signed.
• Ask for the last four weeks of change logs on a comparable account, not a description of the process. Cadence is the thing reviews least reliably capture.
• Ask who supports you — a named team, a shared queue, or a partner. Then ask the reference how long a real escalation took.
• Ask what the platform looked like eighteen months ago, and check that against whatever review you are relying on.

None of that is hostile to any vendor. It is the same diligence we would want a prospect to run on Dr. DSP, and we would answer all four.

How to choose, and what we do instead

Decide on the job, not the brand.

• Your growth problem is reach beyond Amazon — streaming, the open web, audio, digital out-of-home, at national scale. The Trade Desk is a serious answer, its independence is a real advantage, and it has the volume behind it. Staff it properly or buy it through an agency who will.
• Your growth problem is whether Amazon display is adding orders you would not otherwise have got. Buy where the clean room is, and buy the measurement design before the media.
• Both are true. Then run both, and make sure one party is responsible for reconciling them rather than each reporting its own success separately.

Our position, stated so you can score it. Dr. DSP is Amazon DSP run as a managed product: Fable 5 working the account daily with operators from Full Circle supervising — a full-service Amazon management company with more than $500M in managed revenue across 100+ brands. You choose the autonomy level, from every change waiting on your approval through to fully autonomous inside agreed guardrails. Every change carries the evidence behind it, a measurement plan, and a rollback trigger. Orbit is included at no extra cost — and Orbit is software, not a DSP, so it is not a substitute for the platform.

What that book looked like across 30 advertisers in July 2026: 6.04x return on ad spend, 78.4 million impressions at a $4.00 CPM, a blended $1.42 cost per click, and a $5.49 cost per acquisition across 57,137 attributed purchases. One month, thirty advertisers, scoped exactly as measured — not a forecast for anyone.

Where we lose to The Trade Desk, plainly: reach beyond Amazon, connected TV at national scale, supply-path engineering, and the ability to be audited from public filings. We publish no price and they publish no price, so neither of us wins on transparency — Skai, which publishes four exact annual tiers, beats both of us there.

Two places to go next. If the fee structure rather than the product is your question, our breakdown of The Trade Desk pricing computes a take rate from their own filings. If display is not yet the constraint and sponsored ads are where the waste is, Dr. PPC publishes its price — $300 a month plus 3% of ad spend, capped, month-to-month — and starts there.